Fix SARIF-formatter severity levels

[4ch1m]

The SARIF formatter output contains information about "severity levels" in two different contexts (which should be handled differently).

• general "rule-based" severity levels
• "error-/match-based" severity levels within the result set

[ssbarnea]

@4ch1m While I am aware that we dump severity in different places, I do not think that is an issue by itself, mainly because once would be the severity of a rule in general and the other one would specific to a particular occurence.

Still, the title of the PR does not mention or refer the exact bug it is addressing. I am asking this as I want to fully understand what was wrong before. Thanks

[4ch1m]

Sure... I'll try to explain.

The current implementation basically is using this function to obtain the severity (for both the general rule and the specific occurrence:

ansible-lint/src/ansiblelint/errors.py

Lines 116 to 123 in 5ca5679

	@functools.cached_property
	def level(self) -> str:
	"""Return the level of the rule: error, warning or notice."""
	if not self.ignored and {self.tag, self.rule.id, *self.rule.tags}.isdisjoint(
	options.warn_list,
	):
	return "error"
	return "warning"

NOTE: The result is "influenced" by options.warn_list.

External tools (IDE plugins, etc.) will use the SARIF output (and the given severity levels) to highlight/annotate Ansible code; which is problematic in many ways.

Example:
I just recently realized that all of my "name[casing]" issues were marked as errors (whereas their severity is actually only MEDIUM).

External tools should be able to give hints/advice about rules in general (with the rules original/general severity) and the actual error/occurrence.

That's why this should be fixed as proposed in this PR. 😄